This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Fix recent fp-int-convert-float128-timode.c failures
On Wed, Nov 30, 2005 at 09:37:16PM +0100, Manfred Hollstein wrote:
> On Wed, 30 Nov 2005, 17:59:14 +0100, Joseph S. Myers wrote:
> > On Wed, 30 Nov 2005, Manfred Hollstein wrote:
> >
> > > On Wed, 30 Nov 2005, 16:26:47 +0100, Manfred Hollstein wrote:
> > > > Hi there,
> > > >
> > > > recent test results show failures similar to the following:
> > > >
> > > > ERROR: gcc.dg/torture/fp-int-convert-float128-timode.c -O0 : syntax error in target selector "target ia64-*-* && lp64" for " dg-xfail-if 5 "" { ia64-*-* && lp64 } { "*" } { "" } "
> > > >
> > > > I'm currently testing the attached patch as a fix. If that succeeds, I'll
> > > > be checking it in under the obvious-fix rule.
> > >
> > > Hmm, running "make check" still produces the same syntax errors; I'm not
> > > sure at all, if "dg-xfail-if" is capable of combining target patterns
> > > with possibly passed flags. Josef, as you put that patch in which
> > > resulted in these syntax errors, are you sure about it?
> >
> > The patch was based on the documentation of target selectors, which
> > doesn't use a formal syntax so it isn't clear exactly when how many levels
> > of braces are needed, but one possibility is to put in an extra layer of
> > braces, i.e. { dg-xfail-if "" { { ia64-*-* && lp64 } } { "*" } { "" } }.
>
> That doesn't work either unfortunately. To be honest, I'm not sure it
> should, though, as the first argument list passed to dg-xfail-if is
> treated as a list of configure-triplets; ie. there's no code which looks
> at compile options/flags. I'll dig further...
This looks like a bug in the testsuite support for selection expressions,
which I wrote, so I'm looking into it as well.
Janis